Astralane / Developers

Chat Completions

A deliberately bounded OpenAI-compatible chat interface, with real SSE forwarding.

Supported request fields

model, text messages, max_tokens or max_completion_tokens, stream, temperature, function tools, tool_choice (auto / none / required), reasoning_effort, n:1, stream_options.include_usage, routing, and zdr. Unsupported fields fail validation rather than being silently stripped.

Function calls

The chat protocol supports function definitions, streamed tool-call argument fragments, assistant tool_calls, and tool result messages. Your client executes tools; Astralane does not execute arbitrary functions.

Example

The public model name stays the same during provider failover. The selected underlying auto model is reported in platform.billing.selected_model.

Example

{
  "model": "auto",
  "messages": [
    {
      "role": "user",
      "content": "Explain a circuit breaker."
    }
  ],
  "max_tokens": 256,
  "stream": true,
  "routing": {
    "policy": "balanced"
  }
}

SDK compatibility

Examples target familiar OpenAI request shapes. This handoff has native HTTP protocol tests, not a completed official SDK certification suite. Run the included SDK smoke tests locally before claiming drop-in compatibility for a specific SDK/version.

Handoff release

Native core and mock HTTP checks have run. Full dependency install, PostgreSQL/Redis integration, production builds, and browser flows still require local verification. See the repository verification report before deployment.